Abstract: A new time series analysis model based on quantitative recursive feature extraction was proposed,  the time series analysis model of network traffic time series was constructed. The false nearest neighbor algorithm and the average mutual information algorithm were used to reconstruct the phase space of the time series of network traffic.The percent of the frequency characteristic points of the total points of the plane was calculated in the recursive figure plane, and the  feature of network traffic time-frequency entropy was extracted, which can effectively reflect the internal structure of flow time series characteristics and change tendency and  achieve the accurate prediction and monitoring of flow.  Simulation results showed that the proposed algorithm can accurately predict the trajectory of the network traffic, and has good anti-interference ability and higher prediction accuracy.